Forwardsens: A Conceptual Flexible Sensor Solution Integrated Into Forestsens To Track Forwarder Loads Independent of Equipment Manufacturer’s On-Board Systems and Fleet Management Solutions

Real-time information about work progress with respect to production output and product location is required for efficient timber supply chains. Within fully-mechanized timber operations, the productivity of the harvester can be well monitored through the harvester protocol, providing information on the assortment, log volume and numbers, as well as the geolocation of pre-bunched logs at enabled on-board GPS. This information already supports the forwarder routing and beyond can support logistical processes if integrated into a machine specific operation management solution such as John Deere’s TimberMatic. However, such solutions are manufacturer specific and from the landing difficult to further integrate into overall logistic systems of third parties, creating information cut-offs, causing difficulties and delays in trucking and mill production planning. ForwardSens is a proximity sensor as an extension to the RoadSens platform developed within the SmartForest project. The platforms share similar hardware components, i.e., stereo-camera, GNSS receiver and antenna, a communication module, and a GPU-powered edge computer. ForwardSens utilizes the stereo-camera to capture images with depth map and point clouds which are geo-referenced using the data stream of the GNSS receiver. A YOLOv8-based computer vision model is trained and used to detect objects of interest, e.g., the logs and forwarder tracks. The 3D points associated with the objects are derived from the point clouds based on the model output masks and further processed for geometric analysis and mapping purposes. For the validation of the functionality of ForwardSens, an empirical study will be conducted in spring 2024 at an operational site in the communal forest of Oslo/Norway. It is scheduled to monitor the timber extraction conducted by a forwarder during a cut-to-length operation over two consecutive days. The ForwardSens platform will be attached on various positions on the machine (e.g. cabin roof, behind windshield, etc.) to draw conclusions about the influence of common limitations of optical detection systems (view angle, sunlight deflections and dirt aggregation) on the visibility of the working area and objects. Since it is a pilot trial, the extraction cycles will be limited to a single assortment of one length, with data of interest being log number and diameter. The captured loads will be processed by the customized cloud application and compared with the manual counted and scaled logs of every individual load at the roadside landing. The accuracy between the computer vision-based load description and the manual measurements should proof the applicability of optical sensors to support the information flow among logistical processes.